| Summary: | 碩士 === 逢甲大學 === 電機工程學系 === 88 === In many visual communication applications, such as mobile or PSTN vediotelephony, vedioconferencing, and video over the inte rnet, requires very low bit-rate image/vedio coding. However, most of the very low bit-rate image/vedio codec yield visually annoying artifacts that highly degrade the perceptual quality of the image and vedio data. Thus, to achieve very low bit-rate compression while still maintaining the very best perceptual qu ality becomes a very important issue in the image/vedio compre ssion community. The postprocessing in the decoder provide an attractive solution in this issue.
A postprocessing algorithm for the coding artifacts removal of wavelet codec is investigated in the research. This postpro cessing is based on a maximum a posteriori (MAP) estimation in corporated with a Gibbs/MRF model. Utilizing this model, MAP estimate is obtained iteratively using mean field annealing. The decoded image is enhanced from coarse to fine scale, where postprocessing at the coarse scale improves the global appeara nce of the image and reduces long-range artifacts, such as ring ing artifacts, while postprocessing at the fine scale preserves the sharpness of the edge. Then we utilize the properties of Human Visual System(HVS) to predict coding artifacts. Finally we postprocess these coding artifacts. The efficiency of the proposed algorithm is indicated in the experimental results.
|
|---|
